Evaluation of analog memory characteristics of HfO₂-based ferroelectric thin films

Abstract This study investigates the analog memory characteristics of Hf₀.₅Zr₀.₅O₂ (HZO) ferroelectric thin films (6.3 nm) for neuromorphic computing by evaluating their ability to distinguish 3-bit digital input sequences. Eight binary patterns (“000” through “111”) were applied to HZO capacitors, and remanent polarization states were analyzed as a function of switching voltage ( V sw ). Optimal analog memory performance was achieved at V sw = 1.4 V (approximately 1.5 × coercive voltage), which allowed discrimination of the most difficult-to-classify patterns “100” and “010.” Near the coercive voltage (0.9 V), insufficient polarization switching prevented distinct state formation, while at saturation voltages, final polarization became dominated by the most recent input, reducing input history influence. These results identify a critical optimization window balancing polarization switching completeness with memory retention, providing design guidelines for HZO-based neuromorphic devices with enhanced pattern recognition capabilities.